<axd-v>thomassgn: well I have pretty extensive remapping so xmodmap would be a bit insufficient. ***jonsger1 is now known as jonsger
<OriansJ`>axd-v: massive key remapping is pretty rare since; dvork, colemak, maltron, Neo and BEPO are selectable options. Things like single key functions are a bunch more common and usually WM specific. However, if you wish to have a completely new keymap, your first option is probably a better path as it will give you the most possible control. <axd-v>OriansJ`: I see, well in the past I just figured out that adding my own layout variant to the bottom of xkb/symbols/us and also as a variant in xkb/rules/evdev.xml let me find it and select it with setxkbmap. Is there a package in guix that adds additional layouts so that I can inherit and make a similar package for my own layout? I have a custom layout inspired by carpalx that moves around modifier keys and adds custom ISO_level_3 <axd-v>stuff. It was easy before, I just can't figure out how to delete the cash for xkb so that my changed fonts in a directory that I link to are seen by setxkbmap. <OriansJ`>well axd-v an example package for keyboard layout you could crib from is kbd-neo. and if I remember correctly the classical method for clearing the xkb cache is to run: sudo rm -rf /var/lib/xkb/; you just need to tweak it to fix your currently used package. <axd-v>OriansJ`: but I don't have an xkb directory in my /var/lib/ , I don't know why, so I can't wipe the cache. I have found this command by now as well, hereby my question about this in guixsd. <OriansJ`>axd-v: if you look in /gnu/sha256sum-of-program/var/lib/, you'll find what you are looking for <axd-v>OriansJ`: my /gnu/store/xxx-setxkbmap-1.3.1/ has only 2 dirs: bin and share. Share only has doc and man. Not var around. <axd-v>OriansJ`: no other package that I would think would be related to these xkb things has a var/lib with xkb cache files. Don't know what to do. ***abbe_ is now known as abbe
<reepca>Hm, so I've got a program that tries including <gst/gl/gl.h> (which is in the gst-plugins-bad package), which eventually ends up trying to include <gst/video/video.h>, which is in the gst-plugins-base package. pkg-config --cflags gstreamer-gl-1.0, however, doesn't include any mention of directories under the gst-plugins-base output. So naturally the program fails to build. Is this an issue with pkg-config? It sounds to me like "give me <reepca>the necessary flags to include X" is failing to do that. But I guess it's also possible that the build process for the program in question is making some assumption somewhere. <rekado_>according to “guix weather” on berlin 0 of the ~100 missing substitutes for a desktop system are queued. <rekado_>on hydra the substitute situation looks much better. <rekado_>It bothers me that I need to download gigabytes of packages for ghc-pandoc. <rekado_>are we certain that all of them are needed? <rekado_>the packages themselves are also pretty large <rekado_>I wonder if maybe they are actually statically linked but retain obsolete references to the input packages. <efraim>Does Haskell do static linking by default? <efraim>You could try reusing the strip-references code used in go to see if they're actually not needed <firewall`>I try to configure 'GUIX_DAEMON_SOCKET' with ssh <firewall`>I have this error: Failed to read private key: /home/... <efraim>soundtoxin: looks like the path is wrong, you want /./home/brad/.../flash_tool, not /path/./flash_tool <roptat>guix build --check doesn't seem to work: it grafts my package and doesn't actually build anything... <efraim>roptat: guix build --no-grafts --check foo <efraim>Same as you'd want for --log-file <soundtoxin>efraim: any chance you can tell me what change has to be made? I'm using an official script released by a company for flashing software to their device. If it's wrong, I guess they screwed up. <soundtoxin>I see the output directory gets weird near the end, but I don't really understand most of the script <efraim>Line 13 is the 'error', looks like it doesn't know about $dirname/$appname <efraim>Did the script come with instructions? <jonsger>Sleep_Walker: I opened a SR for guix to include the opened bug in the change and spec file :) please accept :P <soundtoxin>that's the closest they've given to instructions for this step <efraim>Try renaming the script to remove the .sh extension <soundtoxin>well, there's a thing in the same directory with that name <soundtoxin>I thought maybe running it from an sshfs-mounted folder might cause weirdness, so I moved it somewhere local <soundtoxin>flash_tool exists in the same directory as flash_tool.sh <soundtoxin>brad@kazuki:~/Downloads/FlashToolLinux/ > ./flash_tool <soundtoxin>-rwxr-xr-x 1 brad users 9.0M Apr 23 16:16 flash_tool <efraim>Then I'd try LD_LIBRARY_PATH=lib ./flash_tool <rekado_>I’d guess that the path to the interpreter has to be patched. <rekado_>the problem here is likely that flash_tool is a foreign binary <soundtoxin>neither the LD_LIBRARY_PATH thing or the bash thing worked <soundtoxin>I was worried guix might cause some weirdness, so I had started to do this on another machine <soundtoxin>but that machine was 32bit and I needed a 64bit machine <rekado_>its interpreter is likely /lib64/ld-2.27.so or something like that. <rekado_>you may need to run patchelf on the binary — or build it from source. <rekado_>another way is to figure out which interpreter this binary expects and then to place a compatible one from a glibc package of the same version in the expected location. <soundtoxin>thanks for the tips. I'll start to explore my options <rekado_>java-bouncycastle has a failing test and then seems to just freeze. <rekado_>org.bouncycastle.mail.smime.test.AllTests <rekado_>I wonder why Icedtea 3 pulls in Icedtea 2. <rekado_>maybe there’s a reference in some text file. <rekado_>indeed, a lot of binaries in the icedtea-3 jdk contain references to icedtea-2 <roptat>what does diffoscope mean by "-Links: 2\\n +Links: 1"? <roptat>do we have an example package that removes timestamps from zip files? <rekado_>we have a build system that does this: ant-build-system <rekado_>false alarm about java-bouncycastle; the check phase eventually passes (despite the failing test suite), but it takes a *very* long time, and it remains silent throughout. <rekado_>the occasional Guile bug when running “guix pull” is scary. <Copenhagen_Bram>What's a luks UUID? And what's the target, is that where the luks partition will be mounted, or the label of the device used to find it? <rekado_>just got a report for a bug in a tool using Conda: <rekado_>_conda/envs/__epicseg@1.0/lib/R/bin/exec/R: error while loading shared libraries: libbz2.so.1.0: cannot open shared object file: No such file or directory <civodul>rekado_: heh, so you told them to "guix package -i r"? :-) <rekado_>civodul: I successfully kept my mouth shut. <rekado_>The problem is that the application manages software on its own. <rekado_>so it would require considerable effort to make it work with anything other than Conda. <jackhill>also, what's the recommended way to get access to an older Emacs for testing? <civodul>jackhill: to build the older version you could use --with-source <civodul>or you could define your own package variant somewhere using (package (inherit emacs) (version xxx) ...) <nckx>Copenhagen_Bram: Just a unique identifier for the LUKS container. The Guix manual explains how to obtain it. The target is what the *decrypted* block device at /dev/mapper/$target will be called. You can then mount /dev/mapper/$target wherever you want it with (file-system (device "/dev/mapper/$target") ...). <rekado_>jackhill: for Guix packages that fail to build it is reasonable to submit a bug report to bug-guix@gnu.org. <rekado_>it’s possible that this is only a problem when building with Guix. <rekado_>whoever maintains the package for Guix could then escalate the problem by involving the upstream developers if it turns out to be a general problem. <rekado_>civodul: yeah, it has deployment built in. (This is Galaxy.) <jackhill>civodul: thanks, I think that I'll try the latter. <jackhill>rekado_: oops, I'll file the report shortly. Is there any additional information that I should put in the bug report other than a link to the hydra build log? <nckx>Copenhagen_Bram: Oh, and remember that you'll be typing your passphrase in a US keymap. <rekado_>jackhill: I recommend including a short description of how to reproduce this and a note of the error you are seeing. It’s especially helpful to mention what you’ve tried to fix this (if taht’s what you did). <rekado_>jackhill: oh, and mention the upstream bug report. <civodul>but note that you can get the build log URL with "guix build --log-file" <Copenhagen_Bram>although, maybe it's a bit sad that the US keymap is the default. Why not the UK keymap? Or the Esperanto keymap? <nckx>Just use ABCDE so it sucks for everyone. <Copenhagen_Bram>Were other distros, such as Parabola or Ubuntu, in a state of beta and had weird bugs such as a messy system init, before GuixSD? Do you think GuixSD has a bright future? <rekado_>I can only answer the second question: yes. <nckx>Copenhagen_Bram: Absolutely on both counts. <nckx>Also: both of those distros are just re-spins of others, while GuixSD is doing things that others (apart from NixOS) don't even do in their own way, so the comparison is of limited use anyway. <pkill9_>i wonder what the first distro based on GuixSD will be <vagrantc>i've seen very few distros really do anything very different in the last 15-20 years <vagrantc>one of the things that attracted me to guix <nckx>Yeah. That was weird. Sunday: there are rumours that MS might be buying GitHub! Monday: oh they did. <roptat>help, my openjdk-9 package is not deterministic <roptat>diffoscope says the differences are on the block count of some files that are otherwise identical <nckx>At least now you won't have to send them any more info to apply for a job. <roptat>also there's a file with a timestamp, let's fix that first <roptat>it says Generated by the IDL-to-Java compiler (portable), version "3.2" <civodul>hmm, might be a different thing, then <civodul>i don't think we have patches for the javadoc issues though <rekado_>roptat: unfortunately, that’s expected. <rekado_>a big part of the icedtea packages is not reproducible <jasmijn>Can I put an NFS mount in the file-systems field? <civodul>jasmijn: i'm not entirely sure it works, but you can try <jasmijn>(file-system (device "hostname:/data") (mount-point "/data") (type "nfs")) <jasmijn>But the device is not valid, of course. <jasmijn>There should be a (title 'network-location) or something like that for this to work. <g_bor[m]>roptat, civodul : Thanks for the help regarding the manuals deleted issue. I think the fix civodul proposed is ok to push. <jasmijn>If there is no out-of-the-box support for NFS mounts, does anyone know of a reasonable approach? <g_bor[m]>The javadoc issue was also mentioned at the Reproducible Builds Summit. It is quite complicated, though. <jasmijn>One way to get NFS mounts would be to put a cronjob at boot with "mount -t nfs hostname:/data /data/", but that seems a very unGuixy way to do it. <g_bor[m]>Timestamps are only part of the problem with javadoc, there were some quite sophisticated filesystem ordering issues, that not only resulted in different order of lines, but different strings expanded to relative and absolute paths, so it was not possible to do the sorting as postprocessing... <civodul>jasmijn: so 'guix system' complains about "hostname:/data"? <civodul>if it does, you can always --skip-checks for now <roptat>does anyone know how I can set my keymap in Xorg without breaking my touchpad? <jasmijn>civodul: Yes, indeed, I got "error: device 'hostname:/data' not found: No such file or directory". <roptat>or do we have an alternative login manager? <jasmijn>With --skip-checks it continues, but then I get: <efraim>that might do it, I tried adding 'ibus' to efl so enlightenment can change keyboards but I forgot to add it to dbus services <jasmijn>well... it doesn't mount, more info coming. <roptat>somehow I can left-click things, but not right-click or move the mouse with my touchpad ^^' <rekado_>jasmijn: I have a cifs file system in my configuration. <civodul>roptat: there's another one, yes (never tried) <jasmijn>rekardo_: Could you send me your cifs file-system definition? <g_bor[m]>roptat: Can you give some more info what are you trying to achieve, what did you try, and what is not working? <rekado_>hmm, guix weather says this about berlin.guixsd.org: build rate: -701.47 builds per hour <roptat>g_bor[m]: I'm trying to configure the keyboard layout, so I've added an extra-config to the xorg-configuration-file <roptat>g_bor[m]: that works, but it also breaks the touchpad, so that I can only left-click <jasmijn>(and note that a simple mount command does work) <jasmijn>Thanks, rekado_; this could be useful. <roptat>so how do I remove slim from the list of %desktop-services? <roptat>well actually it won't solve my issue <roptat>because the configuration doesn't support setting the keymap either <g_bor[m]>roptat: My idea would be to have a look at the configuration file generated, and the X logs. Is it possible that something is broken in xorg-config-file? <roptat>actually let me test without setting the keymap :p <roptat>yeah, it works fine without setting the keymap <janneke>i could use some help trying to compile glibc-2.0.1 <civodul>jasmijn: could you use paste.debian.net for instance? pastebin.com blocks Tor users <g_bor[m]>roptat: %destop-services is just a list, so you can use delete. This discussion came up recently, and civodul posted a more robust solution. <jasmijn>But it's not needed anymore, because the issue is solved. <g_bor[m]>roptat: I guess something along the lines to delete if the service name matches... <jasmijn>No, wait, there are still issues.. let me figure it out. <roptat>g_bor[m]: looking at the Xorg.log file, it seems it unloads the synaptics driver for some reason <roptat>and it doesn't when I don't try to change the keymap <roptat>I needed a MatchIsKeyboard "on" in my xorg configuration <roptat>otherwise it tried to use the evdev driver for the touchpad too <g_bor[m]>roptat: Actually changing the X keyboard layout settings comes up here quite frequently, could we expose this functionality in a more easily accessible way? WDYT? <roptat>I'll try to write something in the manual <jasmijn_>My NFS mount configuration: paste.debian.net/1028107 <jasmijn_>But let me reboot in this instance; had the issue that it didn't boot... <jasmijn>efraim: what do you mean check #f? Should I not have written #t? <bavier`>I've had an issue with nfs on Guix where the connection would timeout and somehow cause the wifi connection to drop, which would then break avahi *.local lookups <jasmijn>You're referring to my config entry, right? <efraim>There's another field to perform a fsck <bavier`>I disabled the nfs a while back and haven't had wifi issues since <jasmijn>fsck would be a bit fsck'ed up for an NFS mount... <jasmijn>bavier`: simple alternative for NFS? <bavier`>jasmijn: not ideal. for me, just scp/rsync the files I want <jasmijn>Hmm, right. Then Unison may be a nice tool! <roptat>hm… anthy says "failed to commit changes to dconf: GDBus.Error:org.freedesktop.DBus.Error.ServiceUnknown: The name ca.desrt.dconf was not provided by any .service files" <civodul>well, looks like we're missing a .service file for dbus <civodul>/gnu/store/gczw5llq7dcw2wph7dky98h53q3gnlk8-dconf-0.26.1/share/dbus-1/services/ca.desrt.dconf.service <civodul>maybe you can add: (simple-service 'dconf dbus-root-service-type (list dconf)) <jasmijn>After adding the nfs mount entry, GuixSD doesn't mount anymore in the most recent instant: it hangs indefinitely just after "input: HDA Intel Front Headphone as ...". <jasmijn>Going back to an older config works. <civodul>it could be that host name lookup fails, or that networking isn't up at that point <jasmijn>civodul: Yes, but especially networking is a good thing to check! <jasmijn>No, I did not know that option, so (needed-for-boot? #f)? <jasmijn>Is it #t by default for file-system entries? <roptat>it seems installing dconf in my profile fixed the issue <civodul>roptat: oh yes, it may be that dconf is not a system service but rather a user service, in dbus terms <civodul>ACTION attempts to replace guix-register with (guix store database) <g_bor[m]>civodul, roptat : if this is the case, then should we make dconf a propagated input of anthy? Would that make sense? <jasmijn>Well, I don't actually need the mount at boot, and I didn't set it to #t, so it should be #f, and that should then be alright? <civodul>g_bor[m]: quite a few packages would need it though, and propagated inputs are frowned upon because they "pollute" the user's profile, so not a great solution <jasmijn>Could I say it should wait until networking is loaded? <civodul>jasmijn: yes, i think you can add something to the 'dependencies' field of <file-system> <jasmijn>Thanks, I'll check it out in the docs. <civodul>whatever the outcome is, it'd be interesting to share your experience on help-guix <reepca>this is actually really puzzling... any attempt to #include <gst/gl/gl.h> will fail, even if you include `pkg-config --cflags gstreamer-gl-1.0` on the command line, unless you also include `pkg-config --cflags gstreamer-plugins-base-1.0` on the command line (it has one of the necessary header files, but pkg-config doesn't know that apparently) <g_bor[m]>civodul: It seems, that in order to use anthy one must have dconf in the profile. If propagating is not a good option, can we point the user to also install dconf? For example by a textual message? <g_bor[m]>This would be also useful for example in the case of gpg and pinentry, but that issue is more complicated... <civodul>g_bor[m]: maybe, i'm not sure what would be a good solution <civodul>something that works out of the box would be better of course <jasmijn>civodul: file-system's "dependencies" only accepts a list of <file-system> or <mapped-device> objects, according to the documentation; so it seems that I cannot define a dependency "networking". <jasmijn>.. even though I also don't know how to define a list of <file-system> objects (by device name?), except by embedding them, but that seems strange to me. <g_bor[m]>civodul: Thinking about issues like .xsession, and the pinentry one do you think that it could make sense to come up with some user configuration description, something similar to operating-system, but the user config files would be generated instead? It would be nice, if we could give the users an option to manage their configurations like we can manage system-generations... WDYT? <g_bor[m]>Or do we have a solution to that already? <civodul>jasmijn: you can manipulate <file-system> like any other Scheme object <civodul>back to the issue, it would be easy to allow for non-file-system and non-mapped-device dependencies <civodul>see dependency->shepherd-service-name in (gnu services base) <civodul>basically just add a (? symbol?) clause there <rekado_>the test failure in vigra is confusing. <rekado_>Assertion failed: Sequences differ at index 1 [(-12.0622,-47.6804) != (-12.0622,-47.6804)] <jasmijn>civodul: Not completeley, not right now. I need to study GuixSD more thoroughly when it's earlier in the day :-) <civodul>i get the exact same error on my laptop <g_bor[m]>rekado_: I second to civodul , it is possible that there is a rounding error... <rekado_>so, should we just disable that test then? <g_bor[m]>rekado_: I guess it is safe to that, but reporting the problem upstream would also be nice. <civodul>it would be good to see if others on the intertubes have the same problem 